two address databases are at work and one perhaps superseding the other
switch address database
ageing of 300 seconds
learning on its downstream ports
no learning on the WLan ports (by design)
bridge fdb
learning on switch downstream ports
learning on WLan ports
If the switch address database is superseding the bridge fdb it would explain why clients that move from a Lan port to a WLan port in less than 300 seconds are exhibiting the connectivity issue.
Hi. This happens on my Turris MOX as well. At first the workaround seem to be working, but testing again it doesn’t. I am not sure why it did seem to work before.
Even when WDS AP enabled on all of my APs (as to my surprise I have read that it may be necessary to get the bridging even working at all)
I have three APs:
[Turris MOX]
| L ethernet -- [ Openwrt on TPLink WR1043ND]
L_ ethernet -- [ Openwrt on D-Link DIR 825 ] -- wireless -- [raspberry-pi]
Each device has a linux bridge over eth and wlan devices With my mobile phone I am roaming between them. When I roam from the D-Link to MOX, it seem to not work until some timeout . When roaming from TP-Link to D-Link, it seems OK just handover to new AP and pings to RPi are going no problem.
Could it be related to ARP caching somehow? I only seem to get it when roaming from my other APs to the Turris, and not when I turn on my device in the Turris range.
I’m sure it’s not ARP caching on the main router, because everything works fine switching between other APs.
I just took my laptop and phone in Turris range, the laptop had no issues and the phone was blocked for a while.
The phone was visible in the wireless clients list with its IP address, but could not be pinged. From the phone, none of the other wireless clients could be reached. After a while the phone disconnected from wifi and tried again but didn’t get a DHCP response. Shortly after that it did get a DHCP response and everything worked.
I think the above behavior matches the phone MAC address being cached and any packets for it going to the LAN interface (that it was on before) and not to the WiFi interface.
This sounds very similar to this issue: OMNIA: Vlan on DSA port breaks arp responses (TOS 4.0.5). There is a script that someone posted which monitors the wifi interface and removes stale entries from the lan bridge when a client roams to wifi. That completely solved the issue for me.